Incident Heart Failure Risk Reclassification With Race-Independent Estimated Glomerular Filtration Rate: A National Heart, Lung, and Blood Institute Pooled Cohorts Analysis.
Aug 4, 2023DOI: 10.1016/j.cardfail.2023.07.009Journal ArticleJCF
Written by: Naman S ShettyVibhu ParchaNirav PatelAkhil PampanaRajat KalraAmbarish PandeyPeng LiAlanna A MorrisSumanth D PrabhuGarima AroraPankaj Arora
Chronic kidney disease (CKD) affects about 15% of US adults and is associated with a high risk of heart failure.1–5 An increasing risk of incident heart failure has been demonstrated with decreasing creatinine-based race-dependent estimated glomerular filtration rate (eGFR).3, 6–8 Based on the eGFR equation used, varying estimates of GFR may be generated for the same individual.9 This may, in turn, lead to differences in the predictive value of eGFR for the estimation of heart failure risk based on the type of eGFR equation used.The most widely used creatinine-based eGFR equation includes a race coefficient to account for the differences in the serum creatinine levels in Black and non-Black individuals.9, 10 However, the race-dependent creatinine-based eGFR may overestimate the GFR in Black individuals. This may affect the predictive value of creatinine-based eGFR for incident heart failure in Black individuals.9, 10 Furthermore, the inclusion of race in the creatinine-based eGFR equation may propagate existing race-based health disparities.11, 12 To reduce race-based health disparities, the National Kidney Foundation (NKF) introduced the new race-independent eGFR based on creatinine and cystatin C that approximated the measured GFR more closely compared with the other equations.9, 13 However, the predictive value of race-independent creatinine and cystatin C-based eGFR for incident heart failure in the overall population and within each race group is unknown.In this study, a pooled cohort analysis was conducted to compare the risk of incident heart failure estimated by the new race-independent creatinine and cystatin C-based eGFR and the existing race-based creatinine-based eGFR equations in the overall population and stratified by race.
Methods
Data was obtained from the National Heart, Lung, and Blood Institute’s Biological Specimen and Data Repository Information Coordinating Center. This study combined the participant-level data from the Atherosclerosis Risk in Communities (ARIC), and the Multi-Ethnic Study of Atherosclerosis (MESA) cohorts. Details of the study design for each of the cohorts have been previously described.14, 15 A brief description of the cohorts has been provided in the Supplementary Methods. Participants free from any cardiovascular disease and with cystatin C and creatinine data were included in the study. The first visit of MESA and the fourth visit of ARIC were considered as the baseline visit for this study. Self-identified race was used to stratify the study population into Black and non-Black individuals. Written informed consent was taken at the respective study centers. Ethical oversight was provided by the University of Alabama Institutional Review Board.
The eGFR equations recommended by the 2021 NKF guidelines were used in this study.9, 13 For the primary analysis, the 2021 Chronic Kidney Disease Epidemiology Collaboration (CKD-EPI) race-independent creatinine and cystatin C-based eGFR and 2009 CKD-EPI race dependent creatinine-based eGFR were used. A secondary analysis using the 2009 race-dependent creatinine-based eGFR and the 2021 CKD-EPI race-independent creatinine-based eGFR has been described in the Supplementary Results. Based on eGFR values, 2 strata of low (<60 ml/min/1.73 m2) and normal (≥60 ml/min/1.73 m2) eGFR were created. Due to skewed distribution, eGFR was divided into quartiles for analysis. Creatinine levels were measured using the modified Jaffe method in all the cohorts. Cystatin C levels were measured using the immunonephelometric method.
Age, sex, diabetes mellitus, anti-hypertensive medication use, smoking history, and statin use were self-reported. Body mass index, blood pressure, and total cholesterol were measured. NT-proBNP (N-terminal proB-type natriuretic peptide) was log-transformed to account for its skewed distribution.
The primary outcome of the study was adjudicated incident heart failure over a 10-year follow-up period. Heart failure included both heart failure with reduced ejection fraction and heart failure with preserved ejection fraction. The diagnostic criteria for heart failure varied with each cohort.
In the ARIC cohort, adjudication of heart failure began in 2005. Prior to 2005, heart failure hospitalizations or deaths associated with International Classification of Diseases 9 (ICD-9) codes beginning with 428 and the International Classification of Diseases 10 (ICD-10) code of I50 were defined as incident heart failure. From 2005, heart failure events were adjudicated by trained staff based on clinical signs and investigations.3, 16
The MESA cohort categorized heart failure as definite, probable, and absent. Probable heart failure was defined as physician-diagnosed heart failure with treatment for heart failure. Definite heart failure required an objective diagnostic criterion such as pulmonary edema on chest X-ray, dilated ventricles, poor left ventricular function, or diastolic dysfunction on imaging studies.16,17
All analyses were performed on SAS 9.4 (Cary, NC). Continuous and categorical variables were summarized using median (interquartile range) and frequency (percentage). The Chi-square test and the Wilcoxon rank-sum test were used to compare categorical and continuous data, respectively. Poisson regression was used to calculate the incidence rate of heart failure in the overall population and in Black and non-Black individuals. The proportionality assumption of hazards was assessed using Schoenfeld’s residuals. Multivariable adjusted Cox regression models were used to assess the 10-year risk of incident heart failure with eGFR in the overall population and stratified by race groups. The models were adjusted for age, sex, body mass index, cohort, diabetes mellitus, systolic blood pressure, total cholesterol, statin use, smoking history, anti-hypertensive use, and log-transformed NT-proBNP. To assess the non-linear association of eGFR and the risk of incident heart failure in the overall population and within the subgroup of race and ethnicity, restricted cubic splines with 3 knots (10th, 50th, and 90th percentiles) were used.18, 19 The change in risk estimation for incident heart failure using the creatinine and cystatin C-based eGFR instead of creatinine-based eGFR was assessed by comparing Harrell’s C-statistics values of the respective Cox models. Additional metrics, including the Akaike Information Criterion, Bayesian Information Criterion, log-likelihood values (−2 log-likelihood), and the Brier score, were used to compare the performance of the fully adjusted Cox models with creatinine and cystatin C-based eGFR and cystatin C-based eGFR.
Results
Among the 15,615 individuals without prevalent cardiovascular risk included in this study, 7,029 (45.0%) were males and 3,726 (23.9%) were Black individuals. The median age of the cohort was 62 (57, 68) years. The median creatinine and cystatin C-based eGFR and creatinine-based eGFR was 91.4 (79.4, 102.0) ml/min/1.73 m2 and 84.7 (72.0, 94.7) ml/min/1.73 m2, respectively. The median creatinine and cystatin C-based eGFR was 91.6 (79.9, 102.0) ml/min/1.73 m2 in non-Black individuals and 90.6 (77.5, 102.1) ml/min/1.73 m2 in Black individuals. The median creatinine-based eGFR was 83.9 (71.8, 93.6) ml/min/1.73 m2 in non-Black individuals and 86.8 (72.9, 102.6) ml/min/1.73 m2 in Black individuals. The first, second, third, and fourth quartiles of the creatinine and cystatin C-based eGFR ranged from <79.4 ml/min/1.73 m2, 79.4–91.4 ml/min/1.73 m2, 91.4–102.0 ml/min/1.73 m2, and >102.0 ml/min/1.73 m2. For the creatinine-based eGFR, the first, second, third, and fourth quartiles ranged from <72.0 ml/min/1.73 m2, 72.0–84.6 ml/min/1.73 m2, 84.6–94.7 ml/min/1.73 m2, >94.7 ml/min/1.73 m2. The baseline characteristics of the overall population and stratified by race are described in Table 1.
There were 879 events of incident heart failure in the overall population. The incidence rate of heart failure was 6.1 (95% CI: 5.8–6.6) events per 1,000 person-years. A non-linear increase in the risk of incident heart failure with decreasing eGFR was noted with both eGFR equations. The estimated risk of incident heart failure was higher using the creatinine and cystatin C-based eGFR compared with the creatinine-based eGFR below 90 mL/min/1.73 m2. With the highest quartile of creatinine and cystatin C-based eGFR as the reference group, the multivariable-adjusted risk of incident heart failure was 1.02 (95% CI: 0.80–1.30) in the third quartile, 1.02 (95% CI: 0.80–1.30) in the second quartile, and 1.47 (95% CI: 1.16–1.86) in the first quartile. With the highest quartile of creatinine-based eGFR as the reference group, the adjusted hazard ratio for incident heart failure was 0.90 (95% CI: 0.73–1.12) in the third quartile, 0.96 (95% CI: 0.77–1.20) in the second quartile, and 1.15 (95% CI: 0.93–1.44) in the first quartile.
Among non-Black individuals, there were 617 events of incident heart failure during the study period. The incidence rate of heart failure was 5.6 (95% CI: 5.2–6.1) events per 1,000 person-years. The risk of incident heart failure increased non-linearly with decreasing eGFR below 90 mL/min/1.73 m2 but the increase in risk was lower with the creatinine-based eGFR compared with the creatinine and cystatin C-based eGFR. With the highest quartile of creatinine and cystatin C-based eGFR as the reference group, the multivariable-adjusted hazard ratio for incident heart failure was 1.19 (95% CI: 0.88–1.62) in the third quartile, 1.11 (95% CI: 0.82–1.51) in the second quartile, and 1.58 (95% CI: 1.17–2.12) in the first quartile. With the highest quartile of creatinine-based eGFR as the reference group, the adjusted hazard ratio for incident heart failure was 1.08 (95% CI: 0.83–1.40) in the third quartile, 1.04 (95% CI: 0.79–1.38) in the second quartile, and 1.30 (95% CI: 0.98–1.70) in the first quartile.
Among Black individuals, there were 262 incident heart failure events over the 10-year follow-up period. The incidence rate of heart failure was 7.9 (95% CI: 7.0–8.9) per 1,000 person-years. On assessing the non-linear association of the risk of incident heart failure with eGFR, the creatinine and cystatin C-based eGFR showed a steeper increase in risk below 90 mL/min/1.73 m2 compared with the creatinine-based equation. With the highest quartile of creatinine and cystatin C-based eGFR as the reference group, the adjusted risk for incident heart failure was 0.84 (95% CI: 0.54–1.31) in the third quartile, 0.89 (0.58–1.36) in the second quartile, and 1.34 (95% CI: 0.90–1.99) in the first quartile. With the highest quartile of creatinine-based eGFR as the reference group, the adjusted hazard ratio for incident heart failure was 1.06 (95% CI: 0.72–1.58) in the third quartile, 1.09 (95% CI: 0.73–1.64) in the second quartile, and 1.19 (95% CI: 0.80–1.77) in the first quartile.
Table 2 summarizes the frequency, incidence rate, and hazard ratio of incident heart failure stratified by the quartiles of eGFR.
Table 3 summarizes the hazard ratio for incident heart failure according to eGFR (categorized using 60 mL/min/1.73 m2 as a threshold) in the overall population and stratified by race.
The non-linear association of the risk of incident heart failure with the creatinine and cystatin C-based eGFR and creatinine-based eGFR stratified by race has been depicted in Figure 1.
In the Cox models for incident heart failure, the C-index was 0.8029 (0.7886–0.8172) using creatinine-based eGFR and 0.8046 (0.7903–0.8189) using the creatinine and cystatin C-based eGFR.
Among non-Black individuals, the C-index for Cox models predicting the risk of incident heart failure was 0.8077 (0.7908–0.8246) using creatinine-based eGFR and 0.8090 (0.7921–0.8259) using the creatinine and cystatin C-based eGFR.
Among Black individuals, the C-index for Cox models predicting the risk of incident heart failure was 0.7944 (0.7674–0.8214) using the creatinine-based eGFR and 0.7976 (0.7709–0.8243) using the creatinine and cystatin C-based eGFR.
The Akaike Information Criterion, Bayesian Information Criterion, log-likelihood values, and the Brier score were similar for the Cox models using creatinine-based eGFR and creatinine and cystatin C-based eGFR in the overall population and stratified by race. (Table 4)
Discussion
In this pooled cohort study of ~16,000 individuals, the creatinine and cystatin C-based eGFR but not the creatinine-based eGFR had modest risk prediction value for the incident heart even after accounting for NT-proBNP in the Cox models. Though the creatinine and cystatin C-based eGFR predicted the risk of incident heart failure at a higher eGFR compared with the creatinine-based eGFR on assessing the non-linear association of the risk of incident heart failure with eGFR, the effect estimates for the association of incident heart failure with per unit decrease in eGFR were similar using both equations. Using the creatinine and cystatin C-based eGFR instead of the creatinine-based eGFR in the models to assess the risk of incident heart failure did not lead to a meaningful change in C-statistics or other metrics of model performance such as AIC, BIC, log-likelihood values, and the Brier score. To summarize, the new creatinine and cystatin C-based eGFR did not improve risk prediction for incident heart failure compared with the creatinine-based eGFR in the overall population and across racial groups.
Although both creatinine and cystatin C-based eGFR and creatinine-based eGFR have no differences in the population-level risk prediction for incident heart failure but using the creatinine and cystatin C-based eGFR instead of the creatinine-based eGFR may have implications in individual-level treatment allocation for heart failure medications such as SGLT-2i and ARNIs. As noted in this study, the creatinine and cystatin C-based eGFR generates higher estimates compared with the creatinine-based eGFR. The higher GFR estimates of the creatinine and cystatin C-based eGFR may increase the eligibility for life-saving heart failure medications because the initiation of these medications is based on the absolute cut-off of the eGFR values. Drugs with a mortality benefit in heart failure such as SGLT-2i are contraindicated in individuals with an eGFR <25 mL/min/1.73 m2.20 Additionally, these patients may also become eligible for dye-based imaging modalities using the creatinine and cystatin C-based equation. However, healthcare systems across the US have adopted the race-independent creatinine-based eGFR equation which underestimates the GFR in Black individuals. This widely adopted change may limit access to heart failure therapy in Black individuals which may lead to the widening of healthcare disparities. Therefore, widespread use of creatinine and cystatin C-based eGFR should be encouraged to improve access to heart failure therapy.
Though there is uncertainty about the direction of causality between heart failure and CKD, several pathophysiological pathways have been elucidated to explain the risk of heart failure in CKD patients. CKD is associated with a high prevalence of heart failure precursors such as left ventricular hypertrophy (LVH) and atherosclerosis.2, 21 LVH in CKD patients may be explained by hypertension, anemia, and increased systemic vascular resistance.1, 2 Hypertension may be attributed to sympathetic overactivation due to stimulation of the renin-angiotensin-aldosterone-system (RAAS) and reduced activity of renalase (an inhibitor of catecholamines).1, 22, 23 RAAS over-activation and dyslipidemia may lead to endothelial dysfunction by increasing oxidative stress.1, 23 CKD is associated with pro-atherogenic lipid abnormalities such as reduction in HDL cholesterol levels, hypertriglyceridemia, and increased levels of oxidized LDL cholesterol levels.1, 22 Reduction of GFR is associated with an increase in asymmetric dimethylarginine levels. Increased asymmetric dimethylarginine levels are associated with reduced nitric oxide production which may lead to increased systemic vascular resistance and endothelial dysfunction.1, 22, 23 Accelerated vascular calcification secondary to hypercalcemia, hyperphosphatemia, and calcification inhibitor deficiency in CKD has been noted.1, 2, 22 The increased oxidative stress, reduced nitric oxide production, and increased vascular calcification coupled with low-grade inflammation seen in CKD promote atherosclerosis.1
Cystatin C, a ubiquitous protein, has been shown to be a more accurate filtration marker compared with creatinine.9 Unlike creatinine, cystatin C levels are less affected by age, sex, weight, race, muscle mass, physical activity, and nutrition.24–26 The GFR equations using cystatin C and creatinine were found to have an agreement of about 90% with the measured GFR.9 The bias between the measured GFR was lower using the creatinine and cystatin C-based eGFR compared with creatinine-based eGFR in Black and non-Black individuals.9 Furthermore, the biases were similar in Black and non-Black individuals using the creatinine and cystatin C-based eGFR.9 The improvement in agreement with the measured GFR and reduced bias of the creatinine and cystatin C-based eGFR may be attributed to the reduced confounding effect of non-GFR determinants while using cystatin C and creatinine together.27 Hence, the creatinine and cystatin C-based eGFR best approximates the renal function compared with the other eGFR equations. Even though the creatinine and cystatin C-based eGFR provides more accurate estimates of eGFR compared with the creatinine-based eGFR, the is no change in the risk prediction value of eGFR in incident heart failure as evidenced by the stable C-statistic values and model performance metrics on replacing the creatinine-based eGFR with the creatinine and cystatin C-based eGFR in the Cox models.
Prior studies have shown that reduced eGFR, regardless of the renal marker used to predict GFR, is a predictor of incident heart failure.2, 3, 6 Considering the varying estimates of eGFR using different renal markers, a study assessing the association of the difference between the 2012 CKD-EPI cystatin C-based eGFR and 2021 race-free creatinine-based eGFR with incident heart failure showed that individuals with a lower cystatin C-based eGFR compared with creatinine-based eGFR had a higher risk of incident heart failure compared with other permutations.24 A UK Biobank-based study assessing the risk reclassification by adding renal markers to the ARIC model for heart failure risk (without natriuretic peptides) showed that the addition of 2012 CKD-EPI cystatin C-based eGFR to the model led to a greater improvement in the C-statistic compared with the addition of 2009 creatinine-based eGFR to the model.28 The current study differs from the prior literature by assessing, 1) the impact of removing race from the eGFR equations through a direct comparison of the predictive values of the new race-independent creatinine and cystatin C-based eGFR and the most widely used race-dependent creatinine-based eGFR in a racially diverse population, and 2) using natriuretic peptides in the risk prediction models for incident heart failure.
Though race is a social construct, racial differences in serum creatinine levels have been noted. These differences led to the inclusion of a race coefficient in the previous eGFR equations. However, the race-based creatinine eGFR equation led to the overestimation of eGFR in Black individuals which limits their access to CKD therapy.9 The introduction of the new eGFR equations attempts to mitigate this disparity by excluding the race coefficient and providing more accurate estimates of GFR.9
There are several limitations to this study. This study combined cohorts that are demographically and clinically different from each other. These differences were accounted for by including the cohort as a covariate in the multivariate-adjusted Cox models. The eGFR equations may not provide reliable estimates for non-White and non-Black individuals and less precise estimates for Black individuals as the validation cohorts had limited Black individuals and insufficient representation of individuals who were not White or Black.9 Though cystatin C levels are less affected by the non-eGFR determinants, factors such as hypothyroidism, glucocorticoid use, inflammation, adiposity, and smoking are known to affect cystatin C levels.29 The current study did not stratify heart failure based on ejection fraction due to limited data. eGFR has been shown to be a stronger predictor of systolic heart failure compared with diastolic heart failure.30 While the analysis controlled for key clinical factors, there may be unaccounted and unmeasured residual confounding contributing to the incident heart failure risk prediction value of eGFR.
Conclusion
The risk prediction value of new race-independent creatinine and cystatin C-based eGFR and the race-dependent creatinine-based eGFR for incident heart failure were comparable. Widespread use of race-independent creatinine and cystatin C-based eGFR may help mitigate health disparities by improving access to heart failure therapy in previously ineligible individuals.
VP
Dr. Vibhu Parcha
Cardiology · Advanced Heart Failure & TransplantationUniversity of Alabama at Birminghamvibhu